Experience Capital: Rethinking Advantage in the Age of Physical AI

For businesses, these shifts aren’t new: there have always been changes in materials, tools, and opportunities. But those that have continued to see success have sought to understand the underlying components of that opportunity and examine what it changes about their strategy. Successful organizations find a way to use these shifts to accumulate an advantage. In the industrial era, companies built advantage by owning productive assets and developing the expertise to use them. A manufacturer could expand its factories, increase output, and build years of knowledge about materials and production. The digital era gave companies another way to build advantage; platforms could learn from customer activity and apply that knowledge across a market without owning every asset needed to deliver the product.

Now, Physical AI will change how companies accumulate advantage — this time, by enabling robots to learn both from physical work and the people who help them perform it. A worker’s explanation of why a robot failed can be used to develop and test a better response, which can then be incorporated into the software guiding other robots, including those at other sites. Expertise previously accessed through a person’s labor can become reusable capital: knowledge that a company retains and applies across many machines. We call this tested, reusable knowledge experience capital, and its value comes from making subsequent work better or less expensive. Creating it requires connecting actions to operating conditions and outcomes, then establishing where a lesson applies. Experience infrastructure is the combination of systems and practices that preserves this knowledge, tests it, and makes it available for further use.

Early examples show how this can work. In Genentech’s research with Anthropic, Claude encountered a liquid-handling problem and repeatedly attempted a recovery that made it worse. But after scientists explained the physical cause and how to correct it, Claude was able to incorporate their explanation into reusable skills that reduced subsequent errors. Similarly, Path Robotics collects data before, during, and after welding to improve its models, and credits production data from individual welding cells for contributing to model improvements delivered across the fleet. The commercial implication is that a customer can gain welding capacity while contributing experience that improves the provider’s offering elsewhere. In both cases, expertise supplied by people became something the system could apply again.

Physical AI involves more than renewed investment in robots and factories. Scarce manufacturing capacity and critical inputs, including parts of the semiconductor and battery supply chains, remain important. But companies can also build an asset from the experience generated as physical work is performed. Physical work generates reusable knowledge, and companies should consider who benefits from it. Advantage may accumulate with the company that can turn that experience into improvements across many operations, whether it owns the equipment or supplies the technology that directs it.

This changes who can capture the value of operating expertise. A company may pay for a deployment and contribute employees’ knowledge, while a provider can incorporate that employee knowledge into a service sold to many customers. The company gains improved operations; the provider may also gain capabilities that strengthen future deployments, including those of competitors. These returns are not determined by the robot alone. They depend on how machines, people, operating context, evaluation, and partners work together—and on what the company can carry forward as that system changes. Leaders should establish the capabilities they want to build before individual deployments set the direction by default.

The urgency lies in commitments being made now. Equipment purchases, systems integration, and commercial agreements can determine what expertise a company can reuse, what it can take to another provider, and what others can commercialize. Leaders should assess those choices not only against a deployment’s productivity targets, but against their implications for margins, asset utilization, and the cost of future expansion. This paper examines how experience capital is created, where its returns may accumulate, and which decisions leaders should make before capital and supplier commitments limit their options.